US20100034090A1 - Edge Node for a network domain - Google Patents

Edge Node for a network domain Download PDF

Info

Publication number
US20100034090A1
US20100034090A1 US12/514,046 US51404609A US2010034090A1 US 20100034090 A1 US20100034090 A1 US 20100034090A1 US 51404609 A US51404609 A US 51404609A US 2010034090 A1 US2010034090 A1 US 2010034090A1
Authority
US
United States
Prior art keywords
load
congestion
data units
node
load reduction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
US12/514,046
Other versions
US8509085B2 (en
Inventor
Attila Bader
Attila Takacs
Andras Csaszar
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Telefonaktiebolaget LM Ericsson AB
Original Assignee
Telefonaktiebolaget LM Ericsson AB
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Telefonaktiebolaget LM Ericsson AB filed Critical Telefonaktiebolaget LM Ericsson AB
Priority to PCT/EP2006/010820 priority Critical patent/WO2008055534A1/en
Assigned to TELEFONAKTIEBOLAGET LM ERICSSON (PUBL) reassignment TELEFONAKTIEBOLAGET LM ERICSSON (PUBL) ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: BADER, ATTILA, CSASZAR, ANDRAS, TAKACS, ATTILA
Publication of US20100034090A1 publication Critical patent/US20100034090A1/en
Application granted granted Critical
Publication of US8509085B2 publication Critical patent/US8509085B2/en
Application status is Active legal-status Critical
Adjusted expiration legal-status Critical

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/12Congestion avoidance or recovery
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/11Congestion identification
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/26Explicit feedback to the source, e.g. choke packet
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/31Tagging of packets, e.g. discard eligibility [DE] bit
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W36/00Hand-off or reselection arrangements

Abstract

An egress node is proposed for a network domain that comprises at least one ingress node for receiving and routing data units belonging to a plurality of flows into said network domain, a plurality of interior routers for routing said data units through said network domain, and said egress node(s). One or more of the interior routers is arranged to detect whether it is subject to a congestion condition and to congestion mark routed data units if it is subject to said congestion condition. The egress node comprises a control element arranged for performing a congestion handling function that comprises a part for detecting the presence of congestion marked data units arriving at said edge node, and a part for reacting to the detection of the presence of congestion marked data units by invoking a congestion control process. The congestion control comprises sending to said ingress node an instruction for terminating one or more flows to thereby reduce a traffic load, waiting a predetermined period of time and then determining whether congestion marked data units are still arriving at said edge node, and if congestion marked data units are still arriving, repeating the sending and waiting steps.

Description

    FIELD OF THE INVENTION
  • The present application relates to an edge node for a network domain that comprises one or more ingress node(s), a plurality of interior routers and one or more egress node(s), to an interior router of such a network domain, and to corresponding control methods for an edge node and an interior router.
  • BACKGROUND OF THE INVENTION
  • In the field of communications packet-switched transport is widely used. Recently, IP-based transport solutions have received more attention. IP stands for Internet Protocol. The interest in IP is at least partially due to the flexibility and wide deployment of IP technologies. However, communication networks, such as telephone networks, have different characteristics when compared to traditional IP networks, because communication networks typically desire to be simple, inexpensive and to be able to provide a defined Quality of Service (QoS). Due to Quality of Service requirements, traffic congestion control is important in communication networks.
  • One of the concepts for providing QoS in IP networks is resource reservation in routers by signalling along the data path. As an example, the Internet Engineering Task Force (IETF) standardization organisation has specified a signalling protocol called RSVP (Resource ReSerVation Protocol) for making resource reservations in IP routers and for providing integrated services for real-time and nonreal-time traffic in the Internet (see e.g. R. Braden at al.: Resource Reservation Protocol (RSVP)—version 1 “Functional Specification”, RFC2205, September 1997; R. Braden at al.: “Integrated Services in the Internet Architecture: an Overview”, RFC1633, 1994; J. Roclawski: “The Use of RSVP with IETF Integrated Services”, RFC2210, September 1997). In this technology, RSVP signalling messages reserve resources in each router along the data path before sending a real-time traffic flow. Real-time flows are admitted into the network if resources are successfully reserved in each router along the data path.
  • This method requires storing per-flow reservation states in each router along the data path. The reservation states are soft states, which means that they have to be refreshed by sending periodic refresh messages. If a reserved state is not refreshed, the state and corresponding resources are removed after a time-out period. Reservations can also be removed by explicit tear-down messages. With the help of per-flow states, RSVP messages always follow the data path up- and downstream. Therefore, it is able to inter-work with standard routing protocols. If the traffic is re-routed, e.g. due to overload of a router, refresh messages make reservations in the new data path. Storing per-flow states in each router also enables the possibility of local path repair. After re-routing, if there are not enough resources to support all flows desired for the new path, some flows can be terminated.
  • Storing and maintaining per-flow states in each router can be a problem in large networks, where the number of flows and therefore the number of reservation states can be high. In response to this scalability problem of RSVP, the IETF specified an RSVP aggregation method that allows making reservations for aggregated flows (see RFC3175, September 2001). Aggregated reservation states are not necessarily created, modified or refreshed for each flow request.
  • The RSVP method of aggregation is capable of making an admission control decision only at an aggregated level, which means that all flows that are treated together as an aggregate are terminated if there are not enough resources on a new path after re-routing.
  • As means for providing QoS in large-scale networks, an architecture referred to as Differentiated Services (DiffServ) was proposed, see RFC2475, 1998. In the DiffServ architecture scalability is achieved by offering services on an aggregate rather than per-flow basis, and by arranging the system such that as much as possible of the per-flow state information is kept at edge nodes of the network, and only aggregated state information is kept in the interior nodes. The service differentiation is achieved by using a specific field in the IP header, the so-called Differentiated Services (DS) field. Packets are classified into Per-Hop Behaviour (PHB) groups at the DiffServ edge nodes. Packets are handled in DiffServ routers according to the PHB indicated in the DS field of the message header. The DiffServ architecture does not specify any way for devices outside of the domain to dynamically reserve resources or receive indications of network resource availability. In practise, service providers rely on subscription-time Service Level Agreements (SLAs) that statically define the parameters of the traffic that will be accepted from a customer.
  • DiffServ networks provide QoS for real-time traffic in a scalable way. DiffServ is a static architecture in which traffic is limited at the domain edges, and interior nodes have to be designed with proper dimensioning. There are no dynamic mechanisms for dealing with overload due to re-routing within the domain, such that interior nodes and links need to be over-dimensioned with respect to average load, in order to provide a QoS guarantee up to at least a certain level.
  • The IETF Next Steps In Signalling (NSIS) working group is working on a protocol to cope with new signalling requirements in IP networks, see RFC3726, April 2004. The QoS signalling application protocol of NSIS is similar to RSVP, but it has some additional features, such as supporting different QoS models. One of the QoS models is Resource Management in DiffServ (RMD). RMD defines scalable admission control methods for DiffServ networks. It also includes a congestion control function that is able to terminate a number of flows in a congestion situation, in order to maintain a required QoS for the rest of the flows.
  • An important characteristic of IP networks and some other packet switched networks compared with other technologies used in communications, such as ATM or SDH, is that IP routing protocols automatically adapt to topology changes for example, if a link or node fails, traffic is automatically re-routed. However, a problem can occur in that the new path may not be able to support all of the re-routed flows, e.g. due to a lack of bandwidth. In such a situation congestion management is desirable.
  • RMD defines functions for notifying edge nodes of a DiffServ domain about congestion occurring in interior routers. These notifications are based on “remarking” of data packets in the DS field in the interior routers in proportion to the overload. An edge note can then measure the number of marked packets and send a termination message for some flows, if it is necessary to reduce the traffic load.
  • An algorithm for reacting to congestion is e.g. described in Andràs Csàszàr, Attila Takàcs, Robert Szabo, Vlora Rexhepi and Georgios Karagiannis “Severe Congestion Handling with Resource Management in DiffServ on Demand”, in proceedings of Networking 2002—The Second International IFIP-TC6 Networking Conference, vol. 2345 of INCS, pages 443-454, 2002, Pisa, Italy, or in Lars Westberg, Andràs Csàszàr, Georgis Karagiannis, Adam Marquetant, David Partain, Octavian Pop, Vlora Rexhepi, Robert Szabo and Attila Takàcs “Resource Management in DiffServ (RMD): A Functionality and Performance Behaviour Overview” in proceedings of PfHSN 2002—7th International Workshop on Protocols for High-Speed Networks, vol. 2334 of INCS, pages 17-34, 2002, Berlin, Germany.
  • A problem encountered in such algorithms can be that of over-reaction to a congestion condition by terminating more flows than necessary to decrease congestion. This is also called under-shoot, in that the actual load falls below a desired level due to the over-reaction.
  • WO 2206/052174 describes ways of dealing with the problem of over-reaction. In the described system, the interior routers of a domain mark packets in proportion to the overload, and the signalled overload is stored. In other words, the interior routers or nodes have a form of “memory” for measured overload. This memorized information is then taken into account when marking new packets in the future. Furthermore, the interior routers mark the forwarded data units proportionally to the degree of overload, and the egress nodes count the marked data units, in order to measure the degree of overload.
  • OBJECT OF THE INVENTION
  • The object of the present invention is to provide an improved way of dealing with congestion in network domains comprising one or more ingress nodes, a plurality of interior routers and one or more egress nodes, especially in view of simplifying the interior routers. Preferably, it is desired to also achieve good congestion response handling without provoking under-shoot, i.e. without terminating flows beyond what is necessary.
  • SUMMARY OF THE INVENTION
  • The object is solved by the subject-matter of the independent claims. Advantageous embodiments are described in the dependent claims.
  • According to a basic embodiment of the present invention, an edge node is provided for a network domain that comprises an ingress node for receiving and routing data units belonging to a plurality of flows into the network domain, a plurality of interior routers for routing the data units through the network domain, and at least one egress node for routing data units out of the network domain. At least one of the interior routers is arranged to detect whether it is subject to a congestion condition and to congestion mark routed data units if it is subject to the congestion condition. Note that it is not necessary that every interior router have this functionality, but that it is preferable that each interior router can act in this way. The congestion marking of routed data units can e.g. be accomplished by setting a certain value in a predetermined field of the header of a routed data unit. The edge node of the invention is arranged to act as the egress node and comprises a control element for performing a congestion handling function that comprises a part for detecting the presence of congestion marked data units, and a part for reacting to the detection of the presence of congestion marked data units by invoking a congestion control process. The congestion control process comprises sending to the ingress node an instruction for terminating one or more flows to thereby reduce a traffic load, waiting a predetermined period of time and then determining whether congestion marked data units are still arriving at the edge node, and if congestion marked data units are still arriving, repeating the steps of sending and waiting.
  • By waiting a predetermined period of time before possibly sending a further instruction to reduce traffic load, the delay between sending such an instruction and the instruction actually being put into place and having an effect can be taken into account. In other words, the above described concept recognizes that there is a certain delay between the sending of a load reduction instruction by the egress node, and the point in time when the egress node is capable of measuring a reaction to that instruction. The delay comprises the time that it takes for the instruction to reach the ingress node, the time that the ingress node takes to follow the instruction, and the time that it takes for data units injected into the domain by the ingress node after following the instruction for again reaching the egress node. By not sending further reduction instructions within the delay period, an under-shoot due to excessively reacting to the reception of congestion marked data units can be avoided in a very simple way. Namely, it is not necessary to specifically adapt the interior routers of the domain, i.e. these can be kept simple.
  • In accordance with a preferred embodiment, one or more interior routers are arranged in such a way that they can detect whether a load exceeds a first load threshold or a second load threshold that is higher than the first load threshold, and to generate congestion marked data units comprising a first marker if the traffic load exceeds the first load threshold, and a second marker that is different from the first marker if the traffic load exceeds the second load threshold. The egress node in turn can then be arranged in such a way that its congestion control process comprises setting a load reduction parameter to a first load reduction value in response to detecting the first marker in received data units, and to a second load reduction value in response to detecting the second marker in received data units, where the second load reduction value corresponds to a larger load reduction than the first load reduction value.
  • The advantage of this concept is that an effective congestion control can be achieved with very simple means. Namely, the interior routers only need to be able to distinguish two different load thresholds for incoming traffic, and to accordingly mark the routed data units. It is not necessary that the data units be marked in proportion to the overload or congestion, as it is e.g. possible that the marking occurs in a step-wise way, i.e. once the respective threshold is exceeded, all subsequently routed data units are accordingly marked. In this way, the interior routers can be implemented in a simple way. Naturally, it is nonetheless possible to combine the concept with that a proportional marking, i.e. that the interior node marks only a fraction of the forwarded data units, proportional to the degree of overload or congestion, once a given threshold is exceeded.
  • In any case, by providing the two different load thresholds and arranging the egress node to accordingly instruct a larger or smaller load reduction, an effective congestion response can be provided, without the necessity of having to implement any complicated mechanisms in the interior routers or the egress nodes. For example, it is not necessary to perform proportional marking in the interior routers, nor is it then necessary to perform measurements of the degree of congestion on the basis of such proportionally marked data units in the egress node.
  • It is noted that the provision of two thresholds and two corresponding load reduction values is only an example, as this concept can naturally also be applied with a greater number of thresholds and accordingly associated load reduction values. For example, in addition to the two thresholds just mentioned, a third threshold could be provided indicating a higher load than the second threshold, and accordingly the egress node could be arranged to respond to third markers indicating the third threshold by sending a load reduction instruction corresponding to a third load reduction value that corresponds to a higher load reduction than the second load reduction value.
  • BRIEF DESCRIPTION OF FIGURES
  • FIG. 1 is a schematic representation of a network domain comprising an ingress node, an egress node and a plurality of interior routers;
  • FIG. 2 is a block diagram schematically showing an edge node according to an embodiment of the invention;
  • FIG. 3 is a flow chart showing a basic method embodiment of the present invention;
  • FIG. 4 is a block diagram schematically showing a further embodiment of an edge node of the present application;
  • FIG. 5 is a flow chart of a further method embodiment of the present invention;
  • FIG. 6 is a flow chart of yet another method embodiment of the present invention;
  • FIG. 7 is a block diagram schematically showing an embodiment of an interior router according to an embodiment of the present invention; and
  • FIG. 8 is a flow chart of an embodiment of a method for controlling an interior router.
  • DETAILED DESCRIPTION OF EMBODIMENTS
  • In the following, specific embodiments of the present invention will be described with reference to the figures. The embodiments are only intended to be seen as examples of the general inventive concepts outlined in the claims, such that they have no limiting effect on the disclosure of these inventive concepts. They only serve to provide the skilled person with a better understanding.
  • FIG. 1 schematically shows a network domain 10 that comprises an ingress node 11 for receiving and routing data units belonging to a plurality of flows into the network domain 10. It furthermore comprises a plurality of interior routers 13, 14, 15 for routing the data units through the network domain 10, and an ingress node 12 for routing the data units out of the network domain 10.
  • It is noted that in the present specification and claims, the term “network domain” is used generically to describe any arrangement of nodes having the just described structure, i.e. one or more ingress nodes, a plurality of interior routers, and one or more egress nodes. As such, a network domain can be a network of its own having routers, where the ingress and egress nodes are provided by appropriate gateway servers to other networks, or a network domain can equally be a sub-division of a larger network. Furthermore, in connection with different protocols and different technologies the sub-divisions of data sent over a network can have different names, such as packets, cells, segments, protocol data units, etc., and the term “data unit” as used in the present specification and claims is generic and relates to any such sub-division of a larger data amount.
  • In the example of FIG. 1, it is assumed that a link failure 19 has occurred between interior routers 14 and 15, such that all traffic from ingress node 11 to egress node 12 that used to pass through interior router 14 has to be re-routed via paths 16, 17 and 18 over interior routers 13 and 15. As a consequence of this re-routing, it could occur that an overload or congestion condition occurs in interior router 13.
  • In the embodiment of FIG. 1, it is assumed that one or more interior routers are arranged to detect whether they are subject to a congestion condition, and to accordingly congestion mark the data units routed through them, if a congestion condition is detected. The congestion condition can be defined in any suitable or desirable way, and equally the congestion marking can be done in any suitable or desirable way. For example, the congestion condition can be determined while measuring the data rate of incoming traffic before queuing, and comparing this value to a threshold value. The congestion condition can equally be defined by measuring the length of a buffering queue and comparing this value with a corresponding threshold value. In each case, if the threshold value is exceeded, then it is determined that a congestion condition is present. The marking of routed data units for indicating congestion can e.g. be done by setting appropriate information in the header of the routed data units. As an example, when using IP data units, the DiffServ Code Point (DSCP) or Explicit Congestion Notification (ECN) field in the IP header can be appropriately set to indicate congestion.
  • The procedure for congestion marking data units can be done in any suitable or desirable way, it can e.g. be done proportional to the overload or it can be cone in stepwise fashion. Proportional marking means that a degree of overload is measured, and only a fraction of the forwarded data unit is marked in accordance with that degree. A stepwise marking means that once the threshold is exceeded, all forwarded data units are marked. A third option can be that the interior node marks packets that are above a threshold.
  • FIG. 2 shows a block diagram of a schematic example of an edge node according to an embodiment of the present invention. Edge node 12, which is arranged to act as an egress node, is shown in FIG. 1. The edge node 12 comprises an input buffer 121 for receiving data units 124 from the network domain 10, and an output buffer 122 for holding data units to be forwarded out of network domain 10. The edge node 12 furthermore comprises a control element 123 that is arranged to perform a congestion handling function that comprises a part for detecting the presence of congestion marked data units arriving at the edge node 12, and a part for reacting to the detection of the presence of congestion marked data units by invoking an appropriate congestion control process. The congestion control process comprises sending to the ingress node 11 an instruction 125 for terminating one or more flows to thereby reduce a traffic load, waiting a predetermined period of time Trelax and then determining whether congestion marked data units are still arriving at the edge node 12. If congestion marked data units are still arriving, then the steps of sending an instruction 125 and waiting the time Trelax is repeated.
  • This procedure is shown in FIG. 3, where step S31 determines whether congestion marked data units are arriving, and if this is the case, then steps S32 and S33 are executed, in which a load reduction instruction is sent to the ingress node, and then a time Trelax is waited. Afterwards, it is again determined in step S31 whether congestion marked data units are arriving, and if this is the case, then steps S32 and S33 are repeated. Otherwise, the procedure basically loops back to continue determining whether congestion marked data units are arriving, where step S34 indicates a possibility for ending the procedure, e.g. via an external interrupt.
  • It is noted that the control element 123 shown in FIG. 2 can be provided in any suitable or desirable way. For example, it can be a microprocessor of known kind for network nodes, which is programmed in such a way that it implements the novel concepts of the present invention. As such, the present invention can also be embodied as a computer program comprising program code parts arranged to perform the method of the invention when executed on a programmable network node. The invention can furthermore be embodied as a computer program product, such as a data carrier, comprising a computer program that is arranged to implement the method of the invention.
  • On the other hand, the control element 123 can also be composed of dedicated hardware elements for implementing the congestion handling function, or can comprise a suitable combination of hardware and software.
  • It is furthermore noted that the edge node 12 shown in FIG. 2 only indicates those features relevant for the explanation of the present invention, such that other known elements of network nodes are not shown for the purpose of simplicity.
  • The node reduction instruction sent by the egress node to the ingress node can be arranged in any suitable or desirable fashion. The instruction can for example be sent via dedicated signalling messages between the egress and ingress nodes, where such signalling messages can be sent over the same paths as data units, possibly using the same protocols, such as IP, or such messages can also be sent over dedicated signalling paths and/or using dedicated signalling protocols.
  • The instruction information itself can be chosen in a variety of ways, as is suitable or desirable. For example, it can be a simple binary signal that can assume one of two values, where one value indicates a reduction and the other indicates no reduction. When the ingress node receives the signal indicating reduction, it reduces the traffic by a predetermined value, e.g. the reduction corresponds to a preset data rate amount, and the ingress node then makes a selection among all flows being handled, in order to achieve the desired load reduction. The selection operation can be done in any suitable or desirable way, e.g. by random choice of flows until so many flows have been terminated that the desired reduction is achieved.
  • In the context of the present specification and claims, the term “flow” is understood generically as relating to data units that are identified by a certain source identifier and destination identifier, e.g. a source address and a destination address. The source and destination identifiers can also have further elements, such as port numbers. As a consequence, real-time or TCP data flows are an example of the generic flows referred to in the present application.
  • As indicated above, the instruction sent by the egress node to the ingress node can be simple binary information that can e.g. be coded by a single bit. However, it is preferable that the instruction comprises more information. According to a preferred embodiment the control element 123 is arranged such that the congestion control process furthermore comprises setting a value of a load reduction parameter in response to the detection of congestion marked data units, and the instruction is such that it instructs the ingress node to terminate one or more flows in accordance with the value of the load reduction parameter. In other words, the load reduction parameter differentiates more exactly how much load to reduce. This can be done in a variety of ways. For example, the load reduction parameter could be expressed in terms of a pre-arranged adjustment factor expressing different degrees of load reduction. For example, there could be a scale of reduction values, where a first value corresponds to a small predetermined adjustment A, a second value to a larger adjustment 2×A and a third value to a larger adjustment 3×A, where the instruction contains the respective first, second or third value, and the ingress node is arranged to accordingly reduce load by terminating an appropriate number of flows.
  • More preferably, the control element and congestion control process are arranged such that the load reduction parameter is expressed in terms of an absolute load or a relative load. In other words, the load can e.g. be measured in an absolute value, such as a data rate, and the instruction can accordingly contain a reduction information in terms of an absolute load, e.g. as an absolute data rate value. The load can also be expressed in relative terms, e.g. as a percentage of total load. As an example, a relative load can be indicated as a fraction of the load caused by a number of flows. The ingress node can then terminate flows from among that number, until the total load (e.g. the total data rate provided by the flows) has been reduced by the indicated fraction.
  • According to a preferred embodiment, the control element and congestion control process in the egress node are furthermore such that a flow detection routine is provided for detecting which flows are affected by congestion, and the instruction to the ingress node comprises an indication of flows from which one or more are to be terminated in accordance with the value of the load reduction parameter. In other words, when looking e.g. at the example of FIG. 3, it is preferable that there is a further step between S31 and S32, which comprises detecting the flows that are affected by congestion. This means that after having detected congestion marked data units, the control element and congestion control process in the egress node identifies which flows contain congestion marks, i.e. which flows carry a marking that indicates congestion. The flows are identified by their source and destination indications. For example, this detection routine can comprise examining the data units being received within a certain period of time, and noting all flow identifiers of data units that carry a congestion marking. The thus determined flow identifiers are then sent to the ingress node with the instruction, such that the ingress node can then terminate flows from the indicated group. The decision regarding which flows to terminate can be made in the ingress node or the egress node. If the decision is made in the ingress node, then it is not necessary to provide more information than the identifiers of the flows affected by congestion. However, according to a further embodiment, the egress node is arranged such that its control element and congestion control process also comprise a flow selection routine for selecting which flows to terminate, and where the instruction to the ingress node then comprises an indication of the selected flows. This embodiment is advantageous in that the ingress node control procedure can be kept simple, as the selection process is performed by the egress node.
  • As already specified earlier, the selection process regarding which flows to terminate can be done in any suitable or desirable way. For example, it can be a process of randomly selecting an affected flow, measuring the load of that flow (e.g. in terms of data rate), determining the load reduction when terminating the selected flow, and then judging whether the termination leads to the reduction as desired by the set value of the load reduction parameter. If the desired load reduction is not achieved, then the process of random selection and following determination is repeated, until so many flows have been considered for termination that the desired load reduction is achieved. The thus selected flows are then indicated to the ingress node for actual termination.
  • Naturally, other forms of flow selection are possible, e.g. it is possible that the individual load of flows is measured and the flows are classified according to their load, from highest load to lowest load. Then, the selection process starts by selecting flows with highest load and examining the effect of termination on load reduction, and selecting flows from highest load towards lowest load, until the desired load reduction is achieved. This selection procedure has the advantage that the selection procedure results in as few flows as possible being terminated in order to achieve the desired load reduction.
  • A further criterion for selecting flows to terminate can depend on the flow type. In other words, the selection process can furthermore take into account a flow type identifier (e.g. real-time flow, non-real-time flow) and then make the selection process dependent on the flow type. For example, the selection process can start by only terminating non-real-time flows. The flow type-based selection can also be combined with the above described random selection or selection based on load contribution.
  • It is noted that the flow identification information can also be used as a means of letting ingress nodes know which instructions are to be processed. This will be explained in more detail in the following. In a simple situation, a network domain only has one ingress node. In this case, each of one or more egress nodes sends its load reduction instruction to that ingress node. On the other hand, if a network domain comprises a plurality of ingress nodes, then there are different possibilities for ensuring a reliable communication between the egress nodes sending instructions and the ingress nodes receiving the instructions. One possibility is that the data units flowing through the network domain comprise an indication of the ingress node that injected the data unit into the network domain. In this case, an egress node receiving a congestion marked data unit can not only identify which flow that data unit belongs to, but also from which ingress node it came. Accordingly, load reduction instructions can be sent to the appropriate ingress node based on that information. As an alternative, in a system in which the data units do not carry indications identifying the ingress nodes, or if such information is not to be used, each egress node can send its instructions to all ingress nodes, together with at least an indication of the flows from which one or more are to be terminated, or with an indication of specific flows to be terminated. Each ingress node can then compare the identified flows with flows being handled by itself, in order to determine whether it is the correct recipient of the instruction. If it is, it terminates the appropriate flows, and if it is not, it simply ignores the instruction.
  • Regarding the setting of a value for the load reduction parameter, this can in principle be done in any suitable or desirable way. For example, if the interior routers of the network domain are arranged to proportionally mark data units when a congestion threshold is exceeded, in ways described previously, then the control element and congestion handling function of the egress node can be arranged to determine the degree of congestion or overload by counting the number of marked and unmarked data units of each affected flow, to thereby receive an indication of the degree of overload on the basis of the relationship between marked and unmarked data units. The value of the load reduction parameter can then be set accordingly, e.g. proportional to the measured degree of overload or congestion. As specified in the introduction to the present application, concepts for arranging interior routers of network domains in such a way that they mark data units proportionally to the degree of overload are known from the prior art, such that a further description is not necessary here.
  • According to a preferred embodiment of the present invention, in order to be able to simplify interior routers and also the control procedure in egress nodes arranged for detecting overload situations, the following new concepts are disclosed. One can arrange interior nodes as e.g. shown in the example of FIG. 7, with a control procedure to be explained with reference to FIG. 8.
  • FIG. 7 shows an interior router that comprises an input buffer 710 for receiving data units 720 to be routed, an output buffer 711 for outputting data units 721 that are to be forwarded, and a control element 712 for controlling the operation of receiving and forwarding data units. As in the example of FIG. 2, further known elements of such routers are not shown for the purpose of simplicity.
  • The control element 712 comprises a part 7121 arranged to detect whether a load at the interior router exceeds a first load threshold th1 or a second load threshold th2, where the second load threshold th2 is larger than the first load threshold th1. The load can e.g. be measured in terms of data rate of arriving data units 720 before queuing. The thresholds may be accordingly expressed in absolute terms, i.e. as a load value, or in relative terms, e.g. as a fraction of a predetermined maximum load value. The interior router of FIG. 7 furthermore comprises a part 7122 in the control element 712 for generating congestion marked data units comprising a first marker L1 if the traffic load exceeds the first load threshold th1, and a second marker L2 different from the first marker, if the traffic load exceeds the second load threshold th2. In the example of FIG. 7, it is assumed that the first load threshold th1 but not the second load threshold th2 is exceeded, and therefore the outgoing data units 721 are marked with L1.
  • An example of a control process is shown in FIG. 8. In a first step S81 it is determined whether the load exceeds the second threshold th2. If yes, the outgoing data units are marked with L2, see step S82, and the process loops back to the beginning. If the load th2 is not exceeded, the procedure goes to step S83, where it is determined whether the threshold th1 is exceeded. If this is the case, then the outgoing data units are marked with L1, see step S84, and the procedure loops back to the beginning. If th1 is also not exceeded, then step S85 determines whether the procedure is to be ended, e.g. based on an external interrupt, and if the procedure is to continue, it loops back to the beginning.
  • It is noted that the control method of FIG. 8 is conveniently simple in that the marking of data units is done in a stepwise fashion, i.e. once a given threshold is exceeded, then all outgoing data units are accordingly marked. There is no proportional marking in this example. The control method of FIG. 8 is one example of how to control the interior router shown in FIG. 7. However, it is noted that the interior router of FIG. 7 can also be operated in a way that includes proportional marking, e.g. that element 7121 not only determines the exceeding of threshold of th1 or th2, but at the same time also measures the degree in which the threshold is exceeded, and that element 7122 marks a proportion of the outgoing data units in correspondence to the degree with which the respective threshold is exceeded.
  • In both cases, the edge node acting as an egress node is preferably arranged in the following way. The egress node can be arranged such that the congestion control process comprises setting the load reduction parameter to a first load reduction value LR1 in response to detecting the first marker L1, and to a second load reduction value LR2 in response to detecting the second marker L2. The second load reduction value LR2 corresponds to a larger load reduction than the first load reduction value LR1.
  • An example of this concept is shown in the method embodiment of FIG. 5. In a first step S500, it is determined whether congestion marked data units are present. If not, the system loops back to the beginning via step 510, in which it is determined whether the procedure should end, e.g. on the basis of an external interrupt. If congestion marked data units are present, step S501 determines whether the marking is an L1 marking. If yes, the procedure advances to step 506, in which the load reduction parameter is set to a first value LR1. Then, in step S507, the procedure identifies the affected flows (i.e. which flows carry L1 markings) and selects which flows to terminate, in order to achieve a load reduction in accordance with LR1. Furthermore, a corresponding instruction is then sent to the ingress node. After that, a period Trelax is set in step S508, during which it is waited, i.e. no further reaction to receiving congestion marked data units is undertaken.
  • After the waiting period Trelax has expired, step S509 determines whether congestion marked data units carrying L1 are still arriving. If yes, then the procedure loops back to step S506, in order to repeat the process of setting the load reduction parameter and sending an instruction for reducing the load to the ingress node. If no further data units carrying L1 are arriving, then the procedure loops back to step S500.
  • If in step S501 it is on the other hand determined that the congestion marked data units are carrying L2 markings, then the procedure advances to steps S502 to S504, which are basically identical to steps S506 to S508, except that in step S502 the load reduction parameter is set to the second value LR2, which indicates a larger load reduction than LR1. After step S504, step S505 determines whether data units carrying L2 markings are still arriving, and if this is the case, the system loops back to step S502, in order to repeat the steps of setting the load reduction parameter to LR2 and then sending an appropriate instruction to the ingress node. On the other hand, if no further data units carrying L2 are arriving, the procedure goes to step S509 to determine whether possibly data units carrying L1 are arriving. The significance of this lies in the fact that LR2 is associated with L2, which in turn is associated with threshold th2 that is larger than the threshold th1, which is in turn associated with L1 and LR1. Namely, if the reduction by LR2 given in step S503 leads to a load reduction at an overloaded interior router below th2, it is possible that the load still lies above th1, i.e. that data units marked with L1 are output, such that a procedure for further reducing the load in accordance with steps S506 to S508 should be followed.
  • The presently described embodiment has the following advantages. When the egress node receives congestion marked data units carrying the marker L2, it can be concluded that an overload exceeding th2 is occurring. If data units carrying L1 are arriving, it can be concluded that an overload exceeding th1 is occurring. Based on the knowledge that th2 corresponds to a larger load than th1, a correspondingly larger load reduction can be initiated (i.e. based on LR2), without having to perform any measurements indicating the degree of overload. As a consequence, both interior routers and egress nodes can be simplified in comparison to the prior art.
  • In the following, examples for choosing the first and second load reduction value will be presented.
  • When using the concept of two-level data unit marking, the marking provides information on the degree of load at an interior router, without having to perform specific measurements. Namely, when data units are L1 marked, it is known that the load is between th1 and th2. Moreover, when L2 marking is detected, the load exceeds th2.
  • Th1 and th2 are preferably chosen in such a way that th1 is associated with a slight overload, whereas th2 is associated with a strong overload that requires a radical load decrease. For example, th1 can be set to lie 5% above an optimum load value, whereas th2 can be set to lie 10% above the optimum load value. Naturally, these are only examples, and other selections are possible.
  • As already indicated previously, th1 and th2 can be expressed in any suitable or desirable way. For example, they can be expressed as absolute load values, e.g. as a data rate, or as a relative value, e.g. as a fraction of an absolute upper limit load. It is noted that in accordance with the present invention it is preferable that all interior routers of the network domain in which the present application is applied use the same values of th1 and th2. Due to this, if the different interior routers of the network domain have different capacities, it is preferable to express th1 and th2 as relative values with respect to a given absolute maximum capacity.
  • As already mentioned, if the egress node receives L2 marked data units, then it can be concluded that the load in an interior router exceeds th2. When reducing the load, it is desired to sharply decrease the load, but preferably not below th1, as this would constitute an under-shoot. As a consequence, it is preferred to choose the second load reduction value that is set if L2 data units are received to a value that corresponds to a load reduction that is equal to the difference between the second load threshold th2 and the first load threshold th1. Namely, by reducing the load by this factor, it is ensured that the load will not fall below th1.
  • Another way of achieving the same effect is to choose the second load reduction value in such a way that it corresponds to a load reduction that is equal to the quotient of the difference between the second load threshold th2 and the first load threshold th1 divided by the second load threshold th2, i.e. equal to (th2−th1)/th2. Choosing the second load threshold in this way is preferable if the invention is applied in a network domain having a plurality of egress nodes. To give an example, one can assume that th1 and th2 are expressed as relative values of an absolute maximum load, e.g. as 50% and 60%, respectively. In this case, the second load reduction value can be set to (60−50)/60=⅙. As a consequence, the edge node that selects the flows in order to reduce the load in accordance with the load reduction value will determine the total load of all affected flows (e.g. the total data rate) and then reduce the number of flows such that the total load is reduced by ⅙.
  • When the egress node receives L1 marked data units, it can conclude that the load at an interior router lies between th1 and th2. As a consequence, the first load reduction value should correspond to a smaller load reduction than the second load reduction value, because using the above examples, a load reduction corresponding to th2−th1 would certainly reduce the load below th1, thereby causing under-shoot. The value of the first load reduction value can be chosen in any suitable or desirable way with respect to a second load reduction value, e.g. as a certain fraction like ½, ⅓, ¼, etc.
  • According to a further embodiment, the control element and congestion control process of the egress node are arranged in such a way that if the step of setting the load reduction parameter to the first load reduction value LR1 is repeated more than once, i.e. that after the time Trelax data units carrying L1 are still received, the value of LR1 is consecutively reduced in each repetition. In other words, the congestion control process is such that it comprises a procedure for sending to the ingress node the termination instruction corresponding to a momentary first load reduction value LR1(i) and then waiting the predetermined period of time Trelax, subsequently determining whether congestion marked data units containing the first marker L1 are still arriving, and if the first marker is still detected, setting the load reduction parameter to a new momentary first load reduction value LR1(i+1) that is smaller than the previous momentary first load reduction value LR1(i), and then repeating the steps of sending the instruction and waiting Trelax.
  • This can e.g. be seen from the example of FIG. 6, which is a variation of the method shown in FIG. 5. Steps in FIG. 6 that are identical to corresponding steps in FIG. 5 carry the same reference numeral, such that a renewed description is not necessary. The difference between the example of FIG. 6 and FIG. 5 is that after determining in step S501 that the congestion marked data units carry L1 markers, a value i is set to zero at step S511. Subsequently, in step S512, which corresponds to step S506 of FIG. 5, the load reduction parameter is set to a first value LR1(i)=LR1(0). Then, in step S507 the flows to be terminated are selected and a corresponding instruction is sent. Afterwards, in step S513 i is incremented by 1. Then, in step S508, the waiting period Trelax is started. If after step S509 the method loops back to step S512, a new first load reduction parameter LR1(i)=LR1(1) is set, where in general LR1(i+1)<LR1(i). It is noted that in the method of FIG. 6, there is also a step S514 for resetting the value of i subsequent to step S505 determining that no L2 marked data units are being received.
  • The significance of choosing ever decreasing values of LR1 is that if one knows that the load is between th1 and th2, then each step of reducing the load makes it more likely that one will fall below th1. As a consequence, in order to avoid under-shoot, it is preferable that the load reduction value LR1 be reduced in each successive iteration.
  • The definition for reducing the subsequent values of LR1(i) can be chosen in any suitable or desirable way. For example, LR1(i) can be chosen as LR1(0)/(i+1). According to a preferred embodiment, the setting of the new momentary first load reduction value LR1(i+1) smaller than the previous momentary value LR1(i) is arranged such that the sum of the overall sequence
  • i = 1 n L R 1 ( i )
  • of consecutively set momentary first load reduction value should be smaller than or equal to the second load reduction value LR2. For example, if LR2 is chosen to correspond to th2−th1, then the sum over LR1(i) should preferably be smaller than or equal to th2−th1.
  • In the general concept of the invention and the above described embodiments, a waiting period Trelax is provided, in order to take into account the delay between the sending of a load reduction instruction by the egress node and the egress node being able to actually determine an affect of having sent that instruction. The waiting period Trelax can be chosen in any suitable or desirable way. For example, it can be set to a fixed value, i.e. when the edge node acting as an egress node is installed, then Trelax is set to a fixed value, e.g. chosen on the basis of measured delay times between edge nodes in the network domain under consideration. As an example, if the average propagation delay between edge nodes in the network domain is T, then Trelax can be set to a value somewhat larger than 2T, e.g. 2.5T.
  • According to a preferred embodiment, the predetermined period of time Trelax is an adaptable value and the control element and the congestion response procedure of the egress node is arranged such that an adaptation routine for the time Trelax is provided. This adaptation routine preferably comprises a routine for determining an indication of a data unit propagation time from the ingress node to the egress node, and adapting Trelax on the basis of that indication of the data unit propagation time. The routine for determining an indication may comprise monitoring periodic refresh messages and/or dedicated query/response messages and/or keep-alive signalling messages exchanged between the ingress node and the egress node. This will be explained with respect to examples in the following.
  • As mentioned above, the time Trelax serves to take into account the delay between sending an instruction and being able to notice an affect of that instruction. As a consequence, the value should preferably be adjusted on the basis of the propagation delay or roundtrip time (RTT) associated with exchanging messages between the egress node and the ingress node. It should be noted that the affect of terminating flows at the ingress node will propagate with the data units of the flows that are not terminated, and the time that it takes these date units to reach the egress node can actually be longer than the average propagation time, due to routing and queuing delays along the path through the network domain. As a consequence, the predetermined waiting period Trelax may be set to be significantly longer than the propagation delay, e.g. 3×RTT to 5×RTT.
  • Different approaches may be used to determine the propagation delay or RTT, depending on the nature of the network domain and the protocols and routines that are employed. For example, if a protocol using soft-states is used (like RSVP or the NSIS signalling protocol), then refresh messages are used to maintain reservation states. These messages are typically sent periodically, and they provide a good opportunity to be used for RTT measurements. Similarly dedicated query/response messages can be used to derive a value for RTT. In NSIS, a query message is defined, which can be sent from an egress node to an ingress node. As a response, the ingress node sends a corresponding response message back to the egress node, where the egress node can measure the delay between sending the query and receiving the response, and use this measurement as an estimate of RTT.
  • Another possibility for assessing RTT can be the use of keep-alive signalling messages sent between edge nodes of a network domain. On the basis of such keep-alive signalling messages, each egress node can maintain a table of propagation delay or RTT values between itself and any other edge node.
  • Another approach to updating the value of Trelax uses worst-case estimates of the RTT. Instead of measuring the RTT by signalling messages, one can also use routine data bases of intra-domain routing protocols (such as USPF, IS-IS) if such protocols are being used. With such a data base, the actual path between any pair of edge nodes can be derived, and by accessing the link queue length properties, the worst-case RTT can be calculated.
  • FIG. 4 shows a preferred embodiment of a control element 123. Reference numeral 1231 refers to a part that determines whether congestion marked data units are being received, but only if Trelax is not currently running, i.e. the system is not in a waiting period. Reference numeral 1232 refers to a part for calculating the amount of traffic to terminate, i.e. setting the value of the load reduction parameter if congestion marked data units are detected. Reference numeral 1233 refers to a part for identifying affected flows and selecting which flows to terminate, in order to provide the load reduction in accordance with the load reduction value. Part 1233 can e.g. be associated with an information base on candidate flows for termination, which flows may be grouped according to their ingress node. After the load reduction value has been set, part 1233 selects the necessary number of flows to realize the desired load decrease. Reference numeral 1235 relates to a part for estimating the delay between sending an instruction and being able to measure the affect of the instruction. For example, it can be an RTT estimator using one or more of the above described concepts. Part 1235 may e.g. maintain an information base of RTT estimates as a function of the ingress nodes. Reference numeral 1234 refers to a part for calculating or setting the waiting period Trelax, preferably on the basis of the estimation made by part 1235 and on the basis of the flows selected in part 1233. The time Trelax can e.g. be set to the maximum RTT among the selected flows, i.e. it is determined which ingress node among the ingress nodes associated with the selected flows is the most distant, i.e. has the largest RTT, and this RTT is then set as Trelax. The set value of Trelax can e.g. be counted down in a timer, where the operation of part 1231 is disabled as long as the timer is counting down. It is noted that the selection operation of part 1233 may also be made dependent on RTT estimates held in the information base of element 1235, in such a way that part 1233 selects flows to terminate that have the smallest stored RTT.
  • The elements 1231-1235 of control element 1233 shown in FIG. 4 can be provided as hardware elements, as software elements executed on a programmable processor or as combinations of hardware and software.
  • The above described concepts can be applied in any suitable context of network domains having ingress nodes, interior routers and egress nodes. It is specifically envisioned to apply the invention in the context of network domains making use of concepts described in the introduction, such as differentiated services (DiffServ) or more specifically Resource Management in DiffServ (RMD). The concepts can also be applied in connection with RSVP or the NSIS signalling protocol.
  • Although the present invention has been described with reference to specific embodiments, these are not intended to have any limiting effect, as they only serve to provide the skilled person with a better understanding, and the scope of protection of the invention is defined by the terms of the independent claims. Reference numerals serve to make the claims easier to read, but do not have any limiting effect.

Claims (31)

1. An edge node for a network domain that comprises at least one ingress node for receiving and routing data units belonging to a plurality of flows into said network domain, a plurality of interior routers for routing said data units through said network domain, and at least one egress node for routing said data units out of said network domain, at least one of said interior routers being arranged to detect whether it is subject to a congestion condition and to congestion mark routed data units if it is subject to said congestion condition, said edge node being arranged to act as said egress node and comprising:
a control element arranged for performing a congestion handling function that comprises
a detecting part, for detecting the presence of congestion marked data units arriving at said edge node,
a reacting part for reacting to the detection of the presence of congestion marked data units by invoking a congestion control process that comprises:
a) sending to said ingress node an instruction for terminating one or more flows to reduce a traffic load,
b) waiting a predetermined period of time and then determining whether congestion marked data units are still arriving at said edge node, and
c) if congestion marked data units are still arriving, repeating steps a) and b).
2. The edge node of claim 1, wherein said control element is arranged such that said congestion control process further comprises setting a value of a load reduction parameter in response to the detection of congestion marked data units, and instructing said ingress node to terminate one or more flows to reduce traffic load in accordance with said value of said load reduction parameter.
3. The edge node of claim 2, wherein said control element is arranged such that said load reduction parameter is expressed in terms of one of an absolute load and a relative load.
4. The edge node of claim 2, wherein said control element is arranged such that said congestion control process further comprises a flow detection routine for detecting which flows are affected by congestion, and from which one or more flows are to be terminated in accordance with said value of said load reduction parameter.
5. The edge node of claim 4, wherein said control element is arranged such that said congestion control process furthermore comprises a flow selection routine for selecting which flows to terminate, and an indication of said selected flows.
6. The edge node of claim 2, wherein said congestion marked data units may comprise a first marker or a second marker, said first marker indicating that a traffic load at an interior node exceeds a first load threshold and said second marker indicating that a traffic load at an interior node exceeds a second load threshold corresponding to a higher load than said first load threshold, and said control element is arranged such that said congestion control process further comprises setting said load reduction parameter to a first load reduction value in response to detecting said first marker and to a second load reduction value in response to detecting said second marker, said second load reduction value corresponding to larger load reduction than said first load reduction value.
7. The edge node of claim 6, wherein said control element is arranged such that said second load reduction value corresponds to a load reduction that is equal to the difference between the second load threshold and the first load threshold.
8. The edge node of claim 6, wherein said control element is arranged such that said second load reduction value corresponds to a load reduction that is equal to the quotient of the difference between the second load threshold and the first load threshold, divided by the second load threshold.
9. The edge node of claim 6, wherein said control element is arranged for
d) sending to said ingress node said instruction for terminating one or more flows to reduce a traffic load corresponding to a momentary first load reduction value,
e) waiting said predetermined period of time and then determining whether congestion marked data units containing said first marker are still arriving at said edge node, and
f) if said first marker is still detected, setting said load reduction parameter to a new momentary first load reduction value smaller than the previous momentary value and repeating steps d) and e).
10. The edge node of claim 9, wherein said control element is arranged such that said setting of said new momentary first load reduction value smaller than the previous momentary value is a sum of an overall sequence of consecutively set momentary first load reduction values is smaller than or equal to the second load reduction value.
11. The edge node of claim 1, wherein said control element is arranged such that said predetermined period of time is a fixed value.
12. The edge node of claim 1, wherein said predetermined period of time is an adaptable value and said control element is arranged such that an adaptation routine for said predetermined period of time is provided.
13. The edge node of claim 12, wherein said control element is arranged such that said congestion handling function comprises a routine for determining a data unit propagation time from said ingress node to said egress node, and said adaptation routine is arranged for adapting said predetermined period of time on the basis of a data unit propagation time.
14. The edge node of claim 13, wherein said routine for determining a data unit propagation time comprises monitoring one or more periodic refresh messages, dedicated query and response messages and keep-alive signalling messages exchanged between said ingress node and said egress node.
15. An interior router for a network domain that comprises at least one ingress node for receiving and routing data units belonging to a plurality of flows into said network domain, a plurality of interior routers for routing said data units through said network domain, and at least one egress node for routing said data units out of said network domain, said interior router comprising:
a control element arranged to detect whether a load at said interior router exceeds a first load threshold or a second load threshold higher than said first load threshold, and to generate congestion marked data units comprising a first marker if the traffic load exceeds the first load threshold and a second marker if the traffic load exceeds the second load threshold.
16. A method for controlling an egress node for a network domain that comprises at least one ingress node for receiving and routing data units belonging to a plurality of flows into said network domain, a plurality of interior routers for routing said data units through said network domain, and at least one egress node for routing said data units out of said network domain, at least one of said interior routers being arranged to detect whether the egress node is subject to a congestion condition and to congestion mark routed data units if it is subject to said congestion condition, the method comprising:
detecting the presence of congestion marked data units arriving at said edge node,
reacting to the detection of the presence of congestion marked data units by invoking a congestion control process that comprises
a) a terminating instruction to said ingress node an instruction for terminating one or more flows to reduce a traffic load,
b) waiting a predetermined period of time and then determining whether congestion marked data units are still arriving at said edge node, and
c) if congestion marked data units are still arriving, repeating steps a) and b).
17. The method of claim 16, further comprising
setting a value of a load reduction parameter in response to the detection of congestion marked data units, and arranging said terminating instruction to terminate one or more flows to reduce traffic load in accordance with said value of said load reduction parameter.
18. The method of claim 17, wherein said load reduction parameter is expressed in terms of one of an absolute load and a relative load.
19. The method of claim 17, wherein said congestion control process further comprises a flow detection routine for detecting which flows are affected by congestion, and said terminating instruction comprises an indication of flows from which one or more are to be terminated in accordance with said value of said load reduction parameter.
20. The method of claim 19, wherein said congestion control process furthermore comprises a flow selection routine for selecting which flows to terminate, and wherein said instruction comprises an indication of said selected flows.
21. The method of claim 17, wherein said congestion marked data units may comprise a first marker or a second marker, said first marker indicating that a traffic load at an interior node exceeds a first load threshold and said second marker indicating that a traffic load at an interior node exceeds a second load threshold corresponding to a higher load than said first load threshold, and said congestion control process furthermore comprises setting said load reduction parameter to a first load reduction value in response to detecting said first marker and to a second load reduction value in response to detecting said second marker, said second load reduction value corresponding to larger load reduction than said first load reduction value.
22. The method of claim 21, wherein said second load reduction value corresponds to a load reduction that is equal to the difference between the second load threshold and the first load threshold.
23. The method of claim 21, wherein said second load reduction value corresponds to a load reduction that is equal to the quotient of the difference between the second load threshold and the first load threshold, divided by the second load threshold.
24. The method of claim 21, wherein said congestion control process further
d) sending to said ingress node said instruction for terminating one or more flows to thereby reduce a traffic load corresponding to a momentary first load reduction value,
e) waiting said predetermined period of time and then determining whether congestion marked data units containing said first marker are still arriving at said edge node,
f) if said first marker is still detected, setting said load reduction parameter to a new momentary first load reduction value smaller than the previous momentary value and repeating steps d) and e).
25. The method of claim 24, wherein said setting of said new momentary first load reduction value smaller than the previous momentary value is arranged such that a sum of an overall sequence of consecutively set momentary first load reduction values is smaller than or equal to the second load reduction value.
26. The method of claim 16, wherein said control element is arranged such that said predetermined period of time is a fixed value.
27. The method of claim 16, wherein said predetermined period of time is an adaptable value and said control element is arranged such that an adaptation routine for said predetermined period of time is provided.
28. The method of claim 27, further comprising
said congestion handling function determining a data unit propagation time from said ingress node to said egress node, and
said adaptation routine adapting said predetermined period of time on the basis of said data unit propagation time.
29. The method of claim 28, wherein said routine for determining a data unit propagation time comprises monitoring one or more periodic refresh messages, dedicated query and response messages and keep-alive signalling messages exchanged between said ingress node and said egress node.
30. A method of controlling an interior router for a network domain that comprises at least one ingress node for receiving and routing data units belonging to a plurality of flows into said network domain, a plurality of interior routers for routing said data units through said network domain, and at least one egress node for routing said data units out of said network domain, comprising:
detecting whether a load at said interior router exceeds a first load threshold or a second load threshold higher than said first load threshold, and
generating congestion marked data units comprising a first marker if the traffic load exceeds the first load threshold and a second marker if the traffic load exceeds the second load threshold.
31.-32. (canceled)
US12/514,046 2006-11-10 2006-11-10 Edge node for a network domain Active 2026-11-12 US8509085B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/EP2006/010820 WO2008055534A1 (en) 2006-11-10 2006-11-10 Edge node for a network domain

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US13/964,220 US9258235B2 (en) 2006-11-10 2013-08-12 Edge node for a network domain

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US13/964,220 Continuation US9258235B2 (en) 2006-11-10 2013-08-12 Edge node for a network domain

Publications (2)

Publication Number Publication Date
US20100034090A1 true US20100034090A1 (en) 2010-02-11
US8509085B2 US8509085B2 (en) 2013-08-13

Family

ID=37814334

Family Applications (2)

Application Number Title Priority Date Filing Date
US12/514,046 Active 2026-11-12 US8509085B2 (en) 2006-11-10 2006-11-10 Edge node for a network domain
US13/964,220 Active 2027-01-19 US9258235B2 (en) 2006-11-10 2013-08-12 Edge node for a network domain

Family Applications After (1)

Application Number Title Priority Date Filing Date
US13/964,220 Active 2027-01-19 US9258235B2 (en) 2006-11-10 2013-08-12 Edge node for a network domain

Country Status (9)

Country Link
US (2) US8509085B2 (en)
EP (1) EP2090034B1 (en)
JP (1) JP4768857B2 (en)
AT (1) AT496463T (en)
AU (1) AU2006350511B2 (en)
CA (1) CA2667108C (en)
DE (1) DE602006019784D1 (en)
ES (1) ES2359646T3 (en)
WO (1) WO2008055534A1 (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100211676A1 (en) * 2008-10-15 2010-08-19 National University Corporation Nagoya University Communication apparatus, communication system, and communication method
US20110016209A1 (en) * 2008-01-14 2011-01-20 Tobias Moncaster Network characterisation
US20110131013A1 (en) * 2008-07-31 2011-06-02 Byoung Hoon Lee Ubiquitous monitoring system
US20110170410A1 (en) * 2010-01-11 2011-07-14 Research In Motion Limited Explicit congestion notification based rate adaptation using binary marking in communication systems
US20110170408A1 (en) * 2010-01-11 2011-07-14 Research In Motion Limited Congestion level indication with explicit congestion notification in communication systems
US20110261695A1 (en) * 2010-04-23 2011-10-27 Xiaoming Zhao System and method for network congestion control
US20130028085A1 (en) * 2011-07-28 2013-01-31 Guy Bilodeau Flow control in packet processing systems
US20130079011A1 (en) * 2010-06-10 2013-03-28 France Telecom Traffic load management method, network and device
US20140164641A1 (en) * 2012-12-11 2014-06-12 The Hong Kong University Of Science And Technology Congestion control for data center traffic

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2237496A1 (en) * 2009-03-31 2010-10-06 BRITISH TELECOMMUNICATIONS public limited company Pre-congestion notification (PCN) as a base for admission control and flow termination
JP5953101B2 (en) * 2012-04-26 2016-07-20 一般財団法人電力中央研究所 Communication system, router, communication method, and control program
ES2588503T3 (en) * 2012-08-27 2016-11-03 Itron, Inc. Bandwidth management in an advanced measurement infrastructure
US9634940B2 (en) * 2013-01-31 2017-04-25 Mellanox Technologies, Ltd. Adaptive routing using inter-switch notifications
US9548960B2 (en) 2013-10-06 2017-01-17 Mellanox Technologies Ltd. Simplified packet routing
US9729473B2 (en) 2014-06-23 2017-08-08 Mellanox Technologies, Ltd. Network high availability using temporary re-routing
US9806994B2 (en) 2014-06-24 2017-10-31 Mellanox Technologies, Ltd. Routing via multiple paths with efficient traffic distribution
US9699067B2 (en) 2014-07-22 2017-07-04 Mellanox Technologies, Ltd. Dragonfly plus: communication over bipartite node groups connected by a mesh network
WO2016019169A1 (en) * 2014-07-30 2016-02-04 Futurewei Technologies, Inc. Method and apparatus for reducing response time in information-centric networks
US9894005B2 (en) 2015-03-31 2018-02-13 Mellanox Technologies, Ltd. Adaptive routing controlled by source node
US9647950B2 (en) 2015-05-11 2017-05-09 Ebay Inc. System and method of site traffic control
US9973435B2 (en) 2015-12-16 2018-05-15 Mellanox Technologies Tlv Ltd. Loopback-free adaptive routing
US10178029B2 (en) 2016-05-11 2019-01-08 Mellanox Technologies Tlv Ltd. Forwarding of adaptive routing notifications
US10200294B2 (en) 2016-12-22 2019-02-05 Mellanox Technologies Tlv Ltd. Adaptive routing based on flow-control credits

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020097726A1 (en) * 2000-10-10 2002-07-25 The Regents Of The University Of California Method for maintaining reservation state in a network router and resulting scalable integrated architectures for computer networks
US20040008628A1 (en) * 2002-07-09 2004-01-15 Sujata Banerjee System, method and computer readable medium for flow control of data traffic
US20040042398A1 (en) * 2002-02-28 2004-03-04 Seriqa Networks Method and apparatus for reducing traffic congestion by preventing allocation of the occupied portion of the link capacity and for protecting a switch from congestion by preventing allocation on some of its links
US20050047340A1 (en) * 2003-08-27 2005-03-03 Jozef Babiarz Technique for end-to-end admission control of real-time packet flows

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2001031860A1 (en) * 1999-10-29 2001-05-03 FORSKARPATENT I VäSTSVERIGE AB Method and arrangements for congestion control in packet networks using thresholds and demoting of packet flows
JP2002344505A (en) * 2001-05-15 2002-11-29 Fujitsu Ltd Packet transfer device, path manager and packet transfer system
JP4295066B2 (en) * 2003-11-12 2009-07-15 株式会社エヌ・ティ・ティ・ドコモ Route control device, communication control device, and communication system using them
JP3836835B2 (en) * 2003-11-18 2006-10-25 株式会社エヌ・ティ・ティ・ドコモ Communication system, monitoring device, edge router, and communication method
EP1810463B1 (en) 2004-11-12 2012-01-04 Telefonaktiebolaget LM Ericsson (publ) Congestion handling in a packet switched network domain
JP2006166059A (en) * 2004-12-08 2006-06-22 Nippon Telegr & Teleph Corp <Ntt> Marking method and marking device

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020097726A1 (en) * 2000-10-10 2002-07-25 The Regents Of The University Of California Method for maintaining reservation state in a network router and resulting scalable integrated architectures for computer networks
US20040042398A1 (en) * 2002-02-28 2004-03-04 Seriqa Networks Method and apparatus for reducing traffic congestion by preventing allocation of the occupied portion of the link capacity and for protecting a switch from congestion by preventing allocation on some of its links
US20040008628A1 (en) * 2002-07-09 2004-01-15 Sujata Banerjee System, method and computer readable medium for flow control of data traffic
US20050047340A1 (en) * 2003-08-27 2005-03-03 Jozef Babiarz Technique for end-to-end admission control of real-time packet flows

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110016209A1 (en) * 2008-01-14 2011-01-20 Tobias Moncaster Network characterisation
US8880681B2 (en) * 2008-01-14 2014-11-04 British Telecommunications Public Limited Company Network characterisation
US8065114B2 (en) * 2008-07-31 2011-11-22 Ajou University Industry Cooperation Foundation Ubiquitous monitoring system
US20110131013A1 (en) * 2008-07-31 2011-06-02 Byoung Hoon Lee Ubiquitous monitoring system
US20100211676A1 (en) * 2008-10-15 2010-08-19 National University Corporation Nagoya University Communication apparatus, communication system, and communication method
US8805997B2 (en) * 2008-10-15 2014-08-12 National University Corporation Nagoya University Communication apparatus detecting abnormaility by communicating an effectiveness notification
US20110170408A1 (en) * 2010-01-11 2011-07-14 Research In Motion Limited Congestion level indication with explicit congestion notification in communication systems
US8693320B2 (en) * 2010-01-11 2014-04-08 Research In Motion Limited Congestion level indication with explicit congestion notification in communication systems
US20140247722A1 (en) * 2010-01-11 2014-09-04 Blackberry Limited Congestion level indication with explicit congestion notification in communication systems
US20110170410A1 (en) * 2010-01-11 2011-07-14 Research In Motion Limited Explicit congestion notification based rate adaptation using binary marking in communication systems
US8416690B2 (en) 2010-01-11 2013-04-09 Research In Motion Limited Explicit congestion notification based rate adaptation using binary marking in communication systems
US9351194B2 (en) * 2010-01-11 2016-05-24 Blackberry Limited Congestion level indication with explicit congestion notification in communication systems
US20110261695A1 (en) * 2010-04-23 2011-10-27 Xiaoming Zhao System and method for network congestion control
US20130079011A1 (en) * 2010-06-10 2013-03-28 France Telecom Traffic load management method, network and device
US9307449B2 (en) * 2010-06-10 2016-04-05 France Telecom Traffic load management method, network and device
US20130028085A1 (en) * 2011-07-28 2013-01-31 Guy Bilodeau Flow control in packet processing systems
US9270556B2 (en) * 2011-07-28 2016-02-23 Hewlett Packard Development Company, L.P. Flow control in packet processing systems
US20140164641A1 (en) * 2012-12-11 2014-06-12 The Hong Kong University Of Science And Technology Congestion control for data center traffic

Also Published As

Publication number Publication date
US9258235B2 (en) 2016-02-09
AT496463T (en) 2011-02-15
AU2006350511A1 (en) 2008-05-15
ES2359646T3 (en) 2011-05-25
JP4768857B2 (en) 2011-09-07
US8509085B2 (en) 2013-08-13
EP2090034A1 (en) 2009-08-19
WO2008055534A1 (en) 2008-05-15
US20130322256A1 (en) 2013-12-05
AU2006350511B2 (en) 2011-05-12
JP2010509814A (en) 2010-03-25
CA2667108C (en) 2015-10-20
EP2090034B1 (en) 2011-01-19
DE602006019784D1 (en) 2011-03-03
CA2667108A1 (en) 2008-05-15

Similar Documents

Publication Publication Date Title
Elek et al. Admission control based on end-to-end measurements
US6690646B1 (en) Network capacity planning based on buffers occupancy monitoring
Oueslati et al. A new direction for quality of service: Flow-aware networking
US7599296B2 (en) Method of transmitting data
CA2683501C (en) An automatic policy change management scheme for diffserv-enabled mpls networks
US6839767B1 (en) Admission control for aggregate data flows based on a threshold adjusted according to the frequency of traffic congestion notification
RU2316127C2 (en) Spectrally limited controlling packet transmission for controlling overload and setting up calls in packet-based networks
US7948878B2 (en) Policing networks
US20060215577A1 (en) System and methods for identifying network path performance
US20050135378A1 (en) Service aware policer with efficient handling of in-profile traffic
US20050226149A1 (en) Method of detecting non-responsive network flows
US8891381B2 (en) Path testing and switching
US8107369B2 (en) Congestion and delay handling in a packet data network
US8797867B1 (en) Generating and enforcing a holistic quality of service policy in a network
US20040125815A1 (en) Packet transmission apparatus and method thereof, traffic conditioner, priority control mechanism and packet shaper
US20050007954A1 (en) Network device and method for categorizing packet data flows and loading balancing for packet data flows
US7729347B2 (en) Method and apparatus for fair flow control and congestion avoidance supporting multiple QoS class requirements
DE60318539T2 (en) Network monitoring system that responds to variations in the variant and mean of the packet arrival times
JP4080911B2 (en) Bandwidth monitoring device
EP1227624A2 (en) Qos monitoring system and method for a high-speed diffserv-capable network element
US6646988B1 (en) System, device, and method for allocating excess bandwidth in a differentiated services communication network
US6958998B2 (en) Traffic management in packet-based networks
US7069356B2 (en) Method of controlling a queue buffer by performing congestion notification and automatically adapting a threshold value
US7327681B2 (en) Admission control method in internet differentiated service network
US10164886B2 (en) Route optimization using measured congestion

Legal Events

Date Code Title Description
AS Assignment

Owner name: TELEFONAKTIEBOLAGET LM ERICSSON (PUBL),SWEDEN

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BADER, ATTILA;TAKACS, ATTILA;CSASZAR, ANDRAS;REEL/FRAME:023495/0878

Effective date: 20090507

Owner name: TELEFONAKTIEBOLAGET LM ERICSSON (PUBL), SWEDEN

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BADER, ATTILA;TAKACS, ATTILA;CSASZAR, ANDRAS;REEL/FRAME:023495/0878

Effective date: 20090507

STCF Information on status: patent grant

Free format text: PATENTED CASE

FPAY Fee payment

Year of fee payment: 4